Second messenger analogues highlight unexpected substrate sensitivity of CD38: total synthesis of the hybrid “L-cyclic inosine 5′-diphosphate ribose”

The multifunctional, transmembrane glycoprotein human CD38 catalyses the synthesis of three key Ca2+-mobilising messengers, including cyclic adenosine 5′-diphosphate ribose (cADPR), and CD38 knockout studies have revealed the relevance of the related signalling pathways to disease. To generate inhibitors of CD38 by total synthesis, analogues based on the cyclic inosine 5′-diphosphate ribose (cIDPR) template were synthesised. In the first example of a sugar hybrid cIDPR analogue, “L-cIDPR”, the natural “northern” N1-linked D-ribose of cADPR was replaced by L-ribose. L-cIDPR is surprisingly still hydrolysed by CD38, whereas 8-Br-L-cIDPR is not cleaved, even at high enzyme concentrations. Thus, the inhibitory activity of L-cIDPR analogues appears to depend upon substitution of the base at C-8; 8-Br-L-cIDPR and 8-NH2-L-cIDPR inhibit CD38-mediated cADPR hydrolysis (IC50 7 μM and 21 µM respectively) with 8-Br-L-cIDPR over 20-fold more potent than 8-Br-cIDPR. In contrast, L-cIDPR displays a comparative 75-fold reduction in activity, but is only ca 2-fold less potent than cIDPR itself. Molecular modelling was used to explore the interaction of the CD38 catalytic residue Glu-226 with the “northern” ribose. We propose that Glu226 still acts as the catalytic residue even for an L-sugar substrate. 8-Br-L-cIDPR potentially binds non-productively in an upside-down fashion. Results highlight the key role of the “northern” ribose in the interaction of cADPR with CD38.

cADPR Acts as a principal second messenger, mobilising intracellular calcium [20][21][22][23] . We are interested in exploiting the common intermediate in cADPR formation and hydrolysis by CD38 24,25 using product-like inhibitors. cADPR Analogues have been accessed by either a chemo-enzymatic route, modelled on its biosynthesis from NAD + , or by total chemical synthesis. Chemo-enzymatic routes rely on Aplysia californica cyclase recognising an NAD + analogue as a substrate and cyclising at the desired N1 position. This limits their use, particularly for analogues that are modified at the locus of the forming N1-glycosidic bond [although see 24 ]. However, cADPR itself is unattractive for inhibitor design since it is readily hydrolysed at the unstable N1 link to give the linear adenosine 5′-diphosphoribose (ADPR), itself a calcium-mobilising second messenger [26][27][28] . We previously reported a chemo-enzymatic route to cyclic inosine 5′-diphosphate ribose via its 8-bromo derivative [N1-cIDPR (abbreviated as cIDPR), 2 and 8-Br-cIDPR, 3, Fig. 1] 29 . Chemically and biologically stable (but vide infra), cIDPR and 8-Br-cIDPR both inhibit cADPR hydrolysis by the catalytic domain of CD38 (shCD38; IC 50 of 276 and 158 µM respectively).
Many total synthetic routes to cADPR analogues required considerable modification of the "northern" ribose in order to generate a stable N1-link [30][31][32][33][34] . The difficulty of selective N1-ribosylation meant that this tended to be the site at which structural modifications were carried out to overcome issues with synthetic tractability. We developed modified Vorbrüggen conditions that effect completely stereo-and regiospecific introduction of an acetylated ribose at the N1-position of a protected inosine [35][36][37] . Unlike the chemo-enzymatic or other synthetic routes, this permits further exploration of the structure-activity relationship at the locus of CD38 catalytic activity using the stable cIDPR template.
Crystallography of shCD38 has identified the mechanism by which NAD + is cyclised to cADPR and ADPR 38 . Glu146 is critical in regulating the multi-functionality of CD38-mediated NADase, ADP-ribosyl cyclase and cADPR hydrolysis activities and Glu226 is the catalytic residue, since its mutation essentially eliminates catalytic activity 39 . Crystal structures obtained with shCD38 and cADPR analogues 40,41 suggest that the "northern" ribose monophosphate region is highly conserved. In the catalytic site, cADPR forms two hydrogen bonds through N6 and N7 to Glu146 and interacts with Glu226 through the 2″-and 3″-OH of the "northern" ribose. Taken together, this suggests a critical role for the base and "northern" ribose in the binding of cADPR analogues to CD38, as might be predicted for the locus of both cADPR formation and degradation. We have previously shown that small fragments consisting of only these elements (N1-inosine 5′-monophosphates, N1-IMP), or analogues of cIDPR with the "southern" ribose replaced by a butyl linker, inhibit cADPR hydrolysis by shCD38 with IC 50 values in the low µM region 36,41 . Crystallography of a hydrolysed cADPR analogue revealed that interactions with Glu146 and Glu226 were maintained even after hydrolysis of the N1-glycosidic bond 36 .
cADPR possesses two ribose sugar motifs and both of these are of the D-configuration. Analogues where the ribose is substituted for an alternative sugar have not previously been explored. Therefore, we designed and synthesised the hybrid cIDPR analogue "L-cIDPR", where the unnatural sugar enantiomer L-ribose is employed to generate an analogue that presents the "northern" ribose hydroxyl groups in a different spatial arrangement. We hoped that an L-configuration in an analogue at this site would probe the interaction between the hydroxyl groups of the "northern" ribose and the catalytic residue Glu226 of shCD38, and predicted that modification of this critical interaction might affect binding affinity.
We report here the synthesis of L-cIDPR and its evaluation as an inhibitor of cADPR hydrolysis by shCD38. L-cIDPR is surprisingly able to access the catalytic machinery of shCD38, being hydrolysed at high enzyme concentrations. We also report the synthesis of further base-modified analogues of L-cIDPR and their activity as inhibitors of cADPR hydrolysis by shCD38 and HPLC studies to examine the stability of the novel analogues towards hydrolysis by shCD38.
Inhibition of cADPR hydrolysis by CD38. The ability of compounds 5-7 to inhibit hydrolysis of cADPR by human shCD38 was evaluated by a fluorimetric cycling assay 42 . The analogues inhibit hydrolysis of cADPR in a concentration-dependent manner (Fig. 3). 8-Br-L-cIDPR 6 and 8-NH 2 -L-cIDPR 7 are good inhibitors, with IC 50 values of 7 μM and 21 μM respectively. In contrast, the parent compound, L-cIDPR 5, is a poor inhibitor with IC 50 of 526 μM, Table 1.
The contrasting inhibitory activity upon substitution of the "northern" D-ribose by L-ribose was unexpected. L-cIDPR 5 is a roughly twofold poorer inhibitor of hydrolysis than cIDPR -which could be ascribed to changes within the "northern" ribose hydroxyl stereochemistry, meaning that critical interactions with the binding pocket are not fully realised. This might be anticipated prima facie, but is nevertheless somewhat surprising given the apparent stereochemical insult afforded by the L-sugar to the CD38 enzymatic machinery. Astonishingly, by comparison, 8-Br-L-cIDPR 6 is highly active with an IC 50 of 7 μM. We had already seen that substitution of cIDPR to generate 8-Br-cIDPR improves the inhibitory activity of this analogue from 276 μM to 158 μM 41 ; however, addition of the 8-bromo atom when the "northern" ribose is in the L-configuration generates a 75-fold increase in inhibitory activity (from 526 μM to 7 μM). In our previous studies, the introduction of an 8-NH 2 substituent on the adenine ring resulted in a further improvement of inhibitory activity (approximately 5-fold increase compared to the parent molecule and 3-fold increase compared to the 8-Br substituted analogue). This increase in binding affinity was shown, by modelling and crystallography, to result from an additional interaction with Asp-155 41 . A comparable trend was not observed for the L-ribose cIDPR analogues prepared here, and this may suggest that the binding of these analogues has changed to adapt to the L-ribose substitution. Conformational analysis. In solution, the ribose ring of nucleosides and nucleotides exists in conformational equilibrium between the C2′-endo and C3′-endo forms. As illustrated in Fig. 4A, the particular conformation adopted affects the spatial presentation of the hydroxyl groups and consequently would be expected to affect the interaction of a ligand with the binding pocket. Indeed, the conformation adopted by the southern ribose in cADPR analogues was shown to underpin their activity at the sea urchin cADPR receptor 43 . Using the method established by Altona and Sundaralingham 44 , the ratio of C2′-endo/C3′-endo forms may be mathematically calculated from the observed coupling constants in the 1 H-NMR spectrum.
We used the 1 H-NMR spectra of analogues 5-7 to determine the syn/anti-conformation of the purine base 45 relative to the N9-ribose and to examine the ribose ring pucker of both the N9 and N1 sugars, Table 2.
As would be expected for conformationally restricted, cyclised analogues with two intact riboses, all three analogues adopt a syn conformation. Analysis of the N9-ribose conformation predicted that 5-7 would adopt a primarily C2-endo ring pucker of the "southern" ribose in free solution -matching that of cIDPR. For the "northern" N1-ribose, the three L-ribose analogues are predicted to have a predominantly C3-endo conformation, calculated using the coupling constant between H-1″ and H-2″ whereas cIDPR displays only a singlet for H-1″, suggesting a dihedral angle of 90° and a C3-endo conformation. The effect of the predominant conformation on 2″-and 3″-hydroxyl group orientation is illustrated for cIDPR (Fig. 4B) and L-cIDPR (Fig. 4C).   (B) From 1 H-NMR data, cIDPR (2) in solution is predicted to display a C3-endo configuration in the "northern" ribose and 61% C2-endo configuration in the "southern" ribose; (C) L-cIDPR (5) is predicted to display a 59% C3-endo and 77% C2-endo configuration, respectively.
SCientifiC RepoRTS | 7: 16100 | DOI:10.1038/s41598-017-16388-0 The "northern" ribose anomeric proton of L-cIDPR is shifted downfield by 0.3 ppm compared to cIDPR, suggesting it is more deshielded and the ring protons H-2″-4″ shifted upfield by 0.2 ppm. These changes reflect the different environments occupied by the ring protons and hydroxyl groups as a result of the change in sugar stereochemistry (Fig. 5).

HPLC studies
We recently discovered 36 that high concentrations of shCD38 were in fact able to hydrolyse the cIDPR scaffold, when crystallography of 8-NH 2 -N9-butyl-cIDPR revealed an N1-hydrolysed product in the active site. This was somewhat surprising, as initial studies at lower concentrations had led to the cIDPR template being described as non-hydrolysable 46 . Hydrolysis would however not be predicted to occur at the concentrations found in enzyme assays, or those likely under physiological conditions ( Supplementary Fig. S1, top two panels), so the cIDPR template can still be assumed to be biologically stable. Study of cIDPR analogues under these conditions can thus be used to evaluate the ability of the enzyme to turn over an analogue, and of an analogue to access the CD38 catalytic machinery. Interestingly, cIDPR is also enzymatically hydrolysable by the related Aplysia cyclase under extreme conditions, as revealed by X-ray crystallographic studies. Incubation of 30 mM cIDPR with wild-type Aplysia cyclase showed that cIDPR was clearly hydrolysed to IDPR at the N1 bond. In this instance, both linear IDPR and the cyclic N1-cIDPR are present and close to each other, with cIDPR in an upside-down orientation compared to cIDPR in CD38 (Q Liu, C Moreau, Q Hao, B V L Potter unpublished data, Supplementary Figure S2). Such an upside down binding mode for cIDPR in the Aplysia enzyme had also been observed earlier 47 . However, cIDPR bound in this way is clearly not an active conformation for cleavage in relation to the known catalytic residue and therefore this mode would be likely to lead to enzyme inhibition.
Incubation of 5 (1 mM final concentration) with 4 mg/mL shCD38 [enzyme in heavy excess] was monitored using RP-HPLC. The peak corresponding to 5 (R T = 12.5 min) reduced in intensity over time, alongside the appearance of a new peak (R T = 15.6 min), that was characteristic of an ADPR analogue (Fig. 6, Supplementary  Fig. S1 third panel). No change in the original peak was observed in a parallel control experiment containing no shCD38. The rate of hydrolysis was similar, or slightly faster than that observed for cIDPR (2) under the same conditions. Surprisingly, treatment of analogue 6 under the same conditions, resulted in no hydrolysis, even after 22 hours ( Fig. 6 and Supplementary Fig. S1, lower panel).
The observed hydrolysis of L-cIDPR at high concentrations of shCD38 demonstrates the flexibility of the CD38 catalytic machinery. The inability of CD38 to hydrolyse 8-Br-L-cIDPR perhaps suggests that this ligand  Table 2. Predicted conformations of L-ribose analogues 5-7. adopts a different binding pose, presumably due to its additional 8-Br substituent combined with the change in hydroxyl stereochemistry resulting from introduction of the L-ribose at N1. Such a change in binding mode would help rationalise the observed difference in inhibition of CD38-mediated cADPR hydrolysis. It would seem feasible that 8-Br-L-cIDPR could be an excellent inhibitor of CD38 and not a substrate by virtue of binding in an upside-down fashion. Such a mode for a hybrid "northern" L-ribose analogue would put the "southern" D-ribose in a pseudo-northern ribose position, and the L-ribose in a pseudo-southern ribose position. We know from previous studies that the southern ribose is not essential for CD38 activity and can be replaced 36 . Unfortunately, attempts to crystallise 8-Br-L-cIDPR with CD38 to determine the exact binding pose were not successful.

Molecular Modelling
Docking of the "northern" ribose modified analogues using the modelling software GOLD generated several different poses in the binding site. No pose, by either visual inspection or by comparison of the docking scores, stood out as being obviously better than any of the others. As an alternative approach, the modified ligands were prepared by amending the stereochemistry of the natural "northern" ribose of cIDPR in the published 2PGJ and 3U4H crystal structures. The 8-substituent was also altered accordingly to 8-H, 8-Br, or 8-NH 2 , to generate a set of ligands in both crystal structures. The protein-ligand complexes were minimised and the binding interactions compared. It is important to note that the latter method is limited, in that it assumes that the L-ribose compounds bind to CD38 with the same basic pose as the natural D-ribose compounds and we would not be able to identify any potential completely different binding poses adopted by the L-cIDPR ligands. However, the approach is justified by the fact that L-ribose cIDPR must be able to access the catalytic residue of CD38 to be a substrate. All analogues had similar poses to cIDPR after minimisation, and showed good overlap of the southern ribose sugar and pyrophosphate backbone. The L-configuration "northern" ribose is rotated around its N1-C1 and C4-C5 bonds, placing the furanose ring oxygen in the centre of the macrocycle and directing the two ring hydroxyl groups towards the binding pocket. The 2″-and 3″-hydroxyl group orientation is significantly changed due to the altered "northern" ribose stereochemistry, and without this rotation, they would point towards the centre of the binding pocket and not be able to contribute to binding. Therefore, the rotation of the ribose maximises the possibility of H-bonds to the catalytic residue Glu226 (Fig. 7).
In all the L-ribose ligand complexes we generated, the L-ribose linked hypoxanthine ring is tilted, suggesting that these ligands may have disrupted π-stacking with Trp189 (Fig. 7C). The potential interaction of the L-ribose itself with the catalytic residue Glu226 is very interesting; the altered stereochemistry of the L-configuration ribose appears to prevent binding of both 2″-OH and 3″-OH to this residue meaning that only one H-bond is formed with Glu226 and the 3″-OH. However, the 2″-OH in the L-ribose conformation is correctly aligned to make a new H-bond with Ser193 (Fig. 7C). This suggests that the catalytically active conformation of L-cIDPR that allows scission of the N1-bond in the active site may differ from cIDPR (Fig. 8). Nevertheless, it is anticipated that only relatively minimal movement of the critical residue Glu226 should be required to accommodate an interaction with the L-ribose hydroxyls and facilitation of attack at the glycosidic bond, in comparison to the natural D-isomer. It is worth noting the immense value of L-nucleosides and derivatives such as Lamivudine, approved against HIV and HBV, that have been shown to have profound antiviral activity with many analogues in clinical trials and abundant evidence for their inhibition of and processing by host and viral enzymes 48 .
We have previously predicted and confirmed by crystallography that the introduction of an 8-NH 2 -substituent can introduce an additional H-bond to Asp-155 in the shCD38 binding pocket 41 . For the 8-substituted L-cIDPR analogues our modelling suggested that the tilt of the hypoxanthine ring to accommodate the rotation of the L-configuration N1-ribose does not interfere with the H-bond to Asp-155 (Supplementary Figure S3). These observations give some rationale to the improved affinity of both 8-Br-L-cIDPR and 8-NH 2 -cIDPR compared to L-cIDPR but seem unlikely to be solely responsible for the extent of the difference in inhibitory activity that is observed.

Conclusion
Synthesis of the proposed ligands was facilitated by selective N1-glycosylation, and three hybrid L-ribose cIDPR-based analogues, substituted in the 8-position of adenine were prepared. Such analogues are unlikely to be accessible via a chemo-enzymatic route and illustrate the utility of our recently developed N1-ribosylation strategy to explore modifications to this sensitive region where both cyclisation and hydrolysis takes place. 8-Br-L-cIDPR is a potent (7 μM) inhibitor of cADPR hydrolysis by CD38. 8-NH 2 -L-cIDPR is similar in potency (21 μM), not showing the enhanced activity relative to its parent that has been previously observed with an extra 8-amino group for another analogue series 41 . Most surprising was the comparable lack of inhibitory activity of the parent compound, L-cIDPR, and its hydrolysis by CD38, albeit at high enzyme concentrations, showing that this analogue can still access the catalytic machinery. In contrast, 8-bromo-L-cIDPR was unaffected by high concentrations of CD38. In our past studies with cADPR analogues docking to CD38 we have been able to employ GOLD to reproduce very accurately the observed crystallographic binding modes, both with a compound from  the inosine-based template, 8-amino-cIDPR, and also with another based upon cADPR itself, the carbocylic analogue cADPcR 41 . Docking studies with the L-ribose-modified ligands based on published crystal structures, however, struggled to predict clear preferred poses for such compounds with the catalytic site, highlighting potentially the critical nature of the stereochemical insult afforded by the L-ribose sugar. Adapting the natural ligand in the active site suggested that interactions could be maintained with the catalytic residue Glu226 by twisting the hypoxanthine ring in the binding pocket and this is supported by the substrate activity of L-cIDPR. It seems feasible, however, that the resistance to hydrolysis of the 8-bromo-L-cIDPR could be ascribed to a non-productive binding mode, possibly one in an upside-down configuration. Taken together, the analogues prepared in this study demonstrate that the "northern" ribose is an area that is very sensitive to modification, presents key binding partners for CD38 and offers some surprising insights.

Experimental Section
General. All reagents and solvents were of commercial quality and were used without further purification, unless described otherwise. The human CD38 catalytic domain (shCD38) was expressed and purified as described previously 36,39 . Unless otherwise stated, all reactions were carried out under an inert atmosphere of argon. 1 H, 13 (10). Intermediate 9 (3.56 g, 4.03 mmol) was taken up in MeOH (20 mL) in a pressure tube. The solution was saturated with NH 3 (g) at 0 °C, then stirred at rt for 12H. The solvent was evaporated and the residue purified by column chromatography on silica gel eluting with DCM:Acetone (1:0 → 0:1 v/v) to afford the title compound

N1-[5′′-O-phosphoryl-β-L-ribofuranosyl]-5′-O-[(diphenylthio)phosphoryl]-8-bromoinosine
Modelling of the L-ribose compounds: Method A. Three cIDPR analogues with L-ribose in the "northern" position but differing in the substituent at the 8-position (hydrogen, bromine or amino) were built using the Schrödinger software and docked into the 2PGJ structure using GOLD 50 . The binding site was defined as a sphere of 5 Å radius centred on the centroid of the cIDPR: the centroid of the docked ligand has to lie within this sphere. Each ligand was docked twenty-five times. The best ranked poses of each the ligand were merged with the protein structure and the resulting complexes passed through a minimisation procedure using the Schrödinger software.
Method B. The ligand in the prepared 2PGJ and 3U4H crystal structures was modified so that the "northern" ribose was of the L-configuration. An 8-amino or 8-bromo group was added to the 8-position of the 2PGJ ligand, and the 8-amino group of the 3U4H ligand was either amended to an 8-bromo substituent or a hydrogen. The resulting complexes were put through 500 rounds of minimisation using the Schrödinger software or minimised using the Discovery Studio software from Accelrys. All figures were prepared using PyMOL.
HPLC studies. The solution containing the CD38 catalytic domain was adjusted to the desired concentration using Tris-HCl buffer (20 mM, pH 8) and 50 μL therefrom was added to the inhibitor (0.05 μmole) in an Eppendorf tube at room temperature. At a given time point, a sample of 5 μL was removed and diluted with 95 μL MilliQ water. 10 μL Of this sample was injected directly into the analytical HPLC system (see General Experimental), eluting at 1 mL/min with an isocratic ion-pair buffer: 0.17% (m/v) cetrimide and 45% (v/v) phosphate buffer (pH 6.4) in MeOH.
Data availability. The crystallographic data for 2PGJ and 3U4H are available in the PDB repository (www. pcsb.org). All other data generated or analysed during this study are included in this published article and Supplementary Information file.